"Fee Switch" Design Space & Next Steps

One problem with only enabling certain pools to “test” is you are now changing the properties of very similar pools. You’re very likely to see liquidity leave ETH/USDT 0.30% pool if you enable the fee switch because they can just join ETH/USDC 0.30% pool without the fee switch on.

If you’re going to turn it on for a pool, it needs to be on all comparable pools (same/similar risk profile) if you actually want a proper test.

Disclaimer: I work at Pantera Capital but views expressed are my own

TLDR:

  • You can’t create protocol fees for free
  • Either traders or LPs have to pay, and in practice both will
  • A protocol switch will result in worse net pricing for traders

@guil-lambert’s post is excellent but could be slightly misleading - there isn’t, of course, any way to create protocol fees ‘out of thin air’ without the ‘tax’ being paid by someone.

Guillaume suggests that liquidity reshuffling between fee tiers would create an equilibrium of LP fees per unit of liquidity provided, which is in theory completely correct. However, this equilibrium would either be:

  • a) Lower than the previous equilibrium - i.e., the protocol fee is ‘spread’ across all fee tiers (and ‘sister pools’ in theory) and still paid by LPs, albeit by each pool in a smaller amount;

  • b) The same/similar, but transaction routing moves trades to tiers with worse net pricing as a result of liquidity reshuffling, and so traders pay the fee;

  • c) Some combination of a) and b), which is what would happen in reality.

A key insight from a) is that a lower equilibrium fee switch, i.e., below the 1/10 minimum, could be applied to all LPs by only turning on the fee switch in select pools (as is being suggested for testing).

To think about this differently, the current liquidity distribution is, of course, optimizing for LP revenue, which in turn is optimizing for maximum daily volume per unit of liquidity in each pool, which in turn is optimizing to give traders the best net pricing experience across the pools. Turning on a switch in any fee tier will alter the liquidity distribution and result in sub-optimal/worse net pricing across all tiers for a given pair/collection of ‘sister’ pairs.

The traders who would have been trading in that pool on the margin will then either continue with their trade anyway (i.e., scenario b) or halt their trade entirely (scenario a).

The extent to which scenario a or b occurs will be determined by the price sensitivity of the traders, which in turn will be determined by the availability of strong alternative protocols.

Thank you for this proposal. I agree that we must explore this design space.

However, I think that you are not giving enough credit and consideration to the unique potential of UNI holders passively accruing fees simply by staking UNI. This possibility should neither be written off nor disregarded. Please consider this more carefully. The Uniswap protocol touches just about every compelling asset in cryptocurrency landscape by enabling permissionless liquidity provision. Now imagine staking UNI and passively accruing an entire portfolio of assets (volume-weighted by trading volume on uniswap).

UNI, in this sense, could become the ultimate Web3 “fixed-income” asset. Holders of UNI can passively accrue every asset that trades on the protocol. This is incredibly powerful and unprecedented. Wealth managers across the world would pile into UNI simply to have this exposure, and accrue exposure for their clients to all Web3 assets trading on the Uniswap protocol. Nothing like this has ever been done before. It’s magical, in a way. Eventually, we can assume that tokenized equities, fixed income assets, tokenized real estate, tokenized fiat currencies, tokenized derivatives/options/swaps/etc and much more may trade on Uniswap v3. And simply by holding the UNI token, one can slowly accumulate exposure to all these different assets… Imagine that. WoW! Incredible…

You say at the beginning of your proposal: “I see an assumption that using the fee switch = money going to UNI token holders. I think this is limiting.”

I disagree strongly. I don’t think this line of thinking is limiting at all. In fact, it is fascinating! I believe this is what UNI was in part destined to be… the primary asset that allows holders to accumulate and accrue a vast and diversified portfolio of positions across every Web3 asset that trades on the Uniswap protocol.

To me, a veteran of traditional finance, this is simply an incredible opportunity for the UNI asset - and it should not be ignored.

2 Likes

So the last community call got me thinking about the fee switch and ways to overcome all the obsticles there are in creating this outside of a contract

The idea of testing this out with V3 having full control over which pools have this fee switch and how much

my proposal is that we create a Staking contract in conjunction of the fee switch contract receiver, and also adding on a ERC20 Wrapped UNI where this wrapped Uni is used in a ETH/WUNI pool, allowing this to be unwrapped as well via contract.

This would allow UNI holders to stake the uni and provide that wrapped UNI to the pool to limit loss from the fees lost in the LP, Also having a UNI/WUNI pool would allow routing through this pool with regular UNI token routes

Now the main reason we can bring in UNI LP providers is that in this contract we would want to retain UNI Holder Voting rights, the Vote would need to be done via this contract and have a ipfs site to vote on, but by tracking and setting the users ability to send vote via the contracts holding of their UNI, this would allow them to still Vote, although a different delegate system would also have to be set up

But by reducing the loss in LP fees and allowing UNI holders to keep their voting power, I do believe this would be enough incentive for LP providers of UNI token at least

Obviously this would be a very complicated contract, but I do believe that this would offer a good incentive to open a Fee Switch Testing pool

3 Likes

Can’t we combine these two things?
Everyone who participated in governance gets 2x as those who only hold UNI tokens but not improve the protocol.

Question is, what about UNI token becoming a security because of this. Also, can we make sure JIT MEV and other scalpers get double taxed ? Or something like that? Do they actually accrue value to the protocol ? Anyway, talk soon.

Wanted to try to give some constructive feedback as a UNIv3 LP and how the fee switch would affect profits and even the decision to LP at all or move liquidity elsewhere.

As a constructive idea that will make this easier for LPs:
The #1 aspect that affects LP profit is trading volume. What if there is a way to have fee switch dialed up/down (0% to 100%) based on a pairs recent trading volume. When volume is down, LP profits are down and the fee switch will be ‘felt’ more for LPs, when volume is up respective to some recent time frame, the fee switch could be dialed up to max and an LP may be fine with a portion of fees being deducted.

Not quite sure how the time frame would work to decide how much to dial up/down fee switch, but just something that could help ease the LPs mind.

2 Likes

What if we added 5-10% to the fees for the end users.

Instead of 0.3% we would charge 0.33% and those that hold a certain amount of Uniswap and have it staked will not have to pay the extra fee + other incentives.

**** Important ****
I would want to see the fees generated being used to also provide liquidity for major pairs acting as a treasury for Uniswap holders where we can compound the rewards in major stable pairs like DAI/USDC further deepening liquidity for the protocol and protecting against those leaving LP.

1 Like

It is better to make data based decisions, then assumptions about the outcome.

A few items that is unknown till data is gathered is:

-Is the liquidity sticky? If a fee switch goes on higher tier pools, do those LP’s move towards a lower tier fee or move to another platform? If the LP’s move to a lower fee tier, then this provides better liquidity, and pricing for traders (instead of “worse net pricing”).

-Is the LP experience, trust and branding on Uniswap enough for LP’s to not mind a fee switch?

I am for having the fee switch on higher fee tier pool’s and for the protocol to collect data on how LP’s in those pool’s react to make more informed decision’s.

I am curious to your insight on how Pantera portfolio companies such as FTX, Bitstamp, Coinbase etc… are able to charge fee’s and make profit without losing trader’s/user’s. Is there a price sensitivity metric of the typical CEX user accross platform’s, or is there some other evaluation tool?

2 Likes

What an exciting conversation!

Our community (Civilization) have been working over the past 12 months on building a decentralised hedge fund, with an initial priority focus on v3 liquidity mining. So we have collected a significant body of evidence, research and technical expertise. Rather intimate knowledge of the mechanisms surrounding concentrated liquidity.

This proposal directly affects and changes our ability to generate a return for our community, and I feel it’s worth sharing my perspective for everyone’s benefit (in case anyone cares to follow up, please feel free to give me a shout @civ100 on Telegram).

In my view, the “fee switch” is equivalent to a “revenues-based income tax”, similar to European-style VAT, whereby the government earns a % of revenues before anything is paid into the seller’s account. There are some technical differences, in who pays how much and how, but substantially the outcome is the same, minus the important “recover VAT on purchases” element, which makes VAT sustainable.

As an income-generating entity, the LP provider will only get 90% of their earnings, whereby 10% is paid to the “government” (UNI governance).

In this way, the “government” effectively can accrue a benefit equivalent to having up to $400mln of TVL invested = 10% of current total TVL of >$4 billion, assuming the fee will be switched on for all pools, which eventually would likely be, post-experiments.

Crucially, the “government” doesn’t suffer any entrepreneurial risk, no impermanent loss, and doesn’t actually need to invest those $400mln.

Is this beneficial, desirable and survivable for both Uniswap and UNI?

Yes, for sure. Uniswap has a strong enough brand and technology lead for sustaining this over the medium term, beyond the initial experiment. The concentrated liquidity research was pioneering and market-defining, and I am sure Uniswap will continue to lead the rest of the DEFI market by a wide margin. However painful this will be for LP, who already contend with extremely razor-thin profits after hedging, this is unavoidable.

Today, v3 is the only protocol with a chance of going mainstream. Also, this must be tested and implemented well ahead of the 2-year code-lock expiring, so it feels like the time has come. Writing and maintaining protocols is complex and expensive, and after all, there must be rewards for their authors, it’s only fair.

But - will this be sustainable long-term? It depends.

If the comparison reference is to the government’s VAT tax, then a 10% tax is relatively cheap when compared to a 20% VAT rate, or even higher in some countries.

If the comparison is the government’s “revenues tax” levied by some countries on e.g. technology companies, then 10% is a lot: 2-3% is more common.

However, crucially, if the comparison was to TradFi, I can’t begin to imagine how exporting Uniswap technology beyond DeFi could sustain this type of monopolistic margins. A traditional market-maker charged 10% of their revenues to access a platform just feels unreal. Market-making will need to be more profitable and easier than would be the case for v3 LPs after the fee switch, so the new technology can be equally beneficial to all parties.

Long story short: I think the fee switch is inevitable, but in Uniswap v4 it better be brought to a lower level than 10%. This brilliant tech will then get the mainstream adoption it fully deserves!

3 Likes

Disclaimer: The authors of this post do not currently hold UNI tokens or provide liquidity to the protocol, although they have in the past. We are an unbiased third party.

This post was co-authored by Matt Fiebach and Dan Smith from Blockworks Research. In our opinion, the protocol fee switch should be activated for a trial period at the minimum 1/10 protocol fee with revenue flowing to the Uniswap treasury address 0x1a9C8182C09F50C8318d769245beA52c32BE35BC.

Uniswap’s most valuable pool is USDC/ETH 0.05% as it single-handedly generates ~50% of protocol volume. Therefore, including it in the trial risks liquidity outflows that jeopardize Uniswap volume dominance.

Selecting the remaining highest volume pools limits the risk of the trial while accurately measuring its impact on LPs. Including one USDC, USDT, and DAI pool allows LPs who are not happy with the fee switch to move to different pools for the same assets.

The following pools should be included in the trial set:

• DAI/ETH 0.05%
• USDT/ETH 0.30%
• USDC/ ETH 1.00%

  1. We agree with “guil-lambert” that LP shuffling is an essential factor in determining which pools to include in the trial set. The repercussion of implementing the protocol fee across the board is very different than for only 1 fee tier as the protocol fee is effectively dispersed across pairs.
  1. Prior to activation, we strongly suggest agreeing on a “circuit breaker” level that discontinues the fee switch if Uniswap stable/ETH volume market share decreases by X% for Y length of time. However, this metric can be maliciously influenced by LPs. If a circuit breaker is not determined prior to the activation of the fee switch, the protocol risks losing market share while governance debates potential options.

It is worth noting that if this test goes successfully without a liquidity drain, it does not necessarily imply that the same would be true if the fee were imposed on all stable/ETH pools.

In the scenario that the trial period is deemed a failure and the fee switch is discontinued (i.e. leads to a negative impact on Uniswap’s stable/ETH volume), we believe LPs should be compensated using treasury held UNI. We propose LPs are repaid, in UNI, for the 1/10 protocol fee they forfeited. To prevent immediate UNI dumping (negative UNI price impact), creating a vesting contract that would look something like LPs getting the protocol fee back in 2 year locked 6 month vested UNI. This essentially acts as a Uniswap treasury swap to diversify assets without the same implications as market selling UNI.

7 Likes

Really happy to see this discussion take place. I’ll share my current thoughts:

Firstly, restating the important and obvious: Currently, all initiatives (costs) are paid by UNI holders by diluting the token price. The only upward pressure to UNI price comes from investors expecting a return in the future. Thus, the first conclusion is that something needs to be done to generate revenue or alternatively run an unsustainable operation to zero.

I’d like to echo @tobyshorin’s first point. I’ve previously had the intuitive sense that activating the fee switch would supercharge Uniswap governance. Toby’s first point argues that having a treasury of stables would make governance participation more attractive. It should also result in more financial discipline in the DAO. The external positive effects seem understated in the discussions.

I believe there is a wide false sense of “having a lot of money to spend” from the UNI treasury (= genesis allocation). This capital should be treated with the same level of rigour as corporate equity. Uniswap needs stables to its treasury to have better understanding of real costs and to always carefully consider when to dilute the UNI token. The fee switch is probably the best bet of accumulating this treasury.

As for the execution, I support @guil-lambert’s direction of testing out the fee switch in various “sister” pools. I do see the risks involved to run the tests on the most important pools as outlined by @BP333.

I’d also encourage people to dive into @rfritsch’s model. The main purpose of running the fee switch test should be to get an understanding of Uniswap’s real market position and possible moat (Modeled by sticky volume & sticky liquidity). We could then use this information to estimate the behavior in other pools.

@MattOnChain’s idea of compensating LPs with vested UNI tokens for possible losses is interesting, but I am against this. Having this type of “bailout” mechanism could distort the test behaviour of LPs, leading to results not beneficial to future use. Furthermore, any protocol revenue generated this way would not count as net revenue. Liquidity mining programs could be run at some point again but should be kept separate from this test.

4 Likes

Actually, they end up charging 0%. Yeah, and they probably have matching engines that charge ultra-low spreads and also do some high-speed trading and so on… On these exchanges, once you have a high enough volume, trading is “free”.

Good writing, however, think about how easy it is to become a market maker with UNI v3, and a profitable one, with each its own parameters.

Taking this into account, 10% is okayish… Not ideal, but okayish, and hell, market makers should buy some UNI if they wish to increase their revenues and have a say in the governance =)

3 Likes

It 's interesting and informative .

It is important. You need to check in practice

Thanks for the feedback!

market makers should buy some UNI

I totally agree with incentivising LPs to buy (more) UNI and directing governance, however, that shouldn’t be the only strategic answer here. Too easily manipulated. So it’s important to have accurate numbers and hard facts to inform the democratic process.

think about how easy it is to become a market maker with UNI v3, and a profitable one, with each its own parameters.

To become a market-maker and earn tokens in liquidity fees is agreeably very easy. Are you sure, however, that to do so on a consistently profitable basis is equally easy?

We have been researching hedging strategies against impermanent loss over quite a long period of time. We ended up with a complex delta-neutral hedge operation, involving automation of multiple dex and cex positions linked to each LP.

This FYI is the main point of our own Civilization protocol - to offer broad access. And if our fund is still in pre-launch mode 1 year later, this is only due to the complexity of an operation that saw other providers attempting a similar development in the past suffering either zero profits, hacks (or both).

For sure, such hedging strategies are beyond reach of the average retail investor. If I remember correctly, even the white paper of Uniswap v3 suggested that devs expected this not to be a retail-friendly system.

Unsurprisingly, some 2021 research suggested that many LPs operate at a loss. Now I do not trust that piece of research at face value, as it was partial and funded by other protocols with conflicting interests.

However, the complexity of running a consistently profitable operation at scale (vis-a-vis buy-and-hold or simple staking) does match my field experience. If my hypothesis was indeed correct, then a 10% tax could make the whole difference between profit and loss, or even sustainable profit vs unsustainably razor-thin margins, even for professional operators.

Do you have personal experience with easily profitable LP strategies on v3 at scale?

If so, I would love to chat further: our decentralised protocol offers immediate incentives to anyone helping to improve our service to the broader DeFi and crypto community.

Making v3 liquidity provision consistently profitable would be a huge step toward mass adoption in TradFi, wouldn’t it? So let’s make it right :slight_smile:

Some people here think that the fee-switch is a standalone 10% fee on each swap. To provide clarity: on the lowest setting, it is a 10% fee on the swap fee.

Asserting my own opinion… I don’t believe taking an academic/research driven approach prior to proceeding with the fee-switch for experimenting with non-core pools is exactly necessary. The Uniswap governance has the flexibility to learn from liquidity data of how toggling the fee switch on the pools suggested by @guil-lambert, and @alanalevin impacts liquidity drain. Backing out of the protocol fee is an option at a later date. I would wager with high conviction that the minimum 10% protocol fee as a percentage of the swap fees on three (small low-TVL) non-core pools is unlikely to be an existential threat to the attractiveness of Uniswap being the dominant AMM for consistent APY.

Only after we have collected data can the community be okay with making heuristic arguments on whether if the protocol fee can be extended to a wider list of pools (or the protocol in its entirety). I don’t mean to sway away researchers, but we need to maintain a healthy balance between practical solutions rather than surgically polished academic research that would realistically take months to produce. Hypothetically speaking, if the protocol fee of 10% on all swap fees can be broadly applied to all pools, what would stop the governance to pushing-up the protocol fee to 25% for all swap fees, or variable based on the classification of the pool? What we’re trying to achieve here can never be fully described and optimized in formal language.

We need to maintain focus and keep the momentum up with a monumental 40M large voting barrier, or we risk stagnation. There probably are going to be times where we can get it wrong for some select pools. Though how hard would it be re-incentivize liquidity by backtracking the fee-switch?

It is not easy if you don’t have the numbers to back it up, it is easy if you have the numbers and stats, to know which type of IL risk you want to take compared to the revenues you should be able to generate. One big caveat though is that the theoretical yields that you can make are in general slashed by MEV players that ripp off a lot of fees with JIT liquidity providing, which at first seems annoying to me. My questoins regarding the MEV JIT LPs is, is it positive for Uniswap? Is it a necessary evil in order to make sure the pools are liquid enough for traders ? Or without it is it better because of a better redistribution to passive LPs ? Maybe UNI should implement a fee switch in order to foster research in this direction, and maybe manage to prevent bad actors (if MEV JIT LPs are actually bad) to “steal” liquidity from “honest” LPs… It’s a lot of question marks here, I honestly don’t have the final answers as to this phenomenon. Any MEV JIT LPs care to answer ? =)

Posted this on twitter but going to repost here because I think this direction is a huge mistake. I absolutely want fee teirs on but turning it on on a few pools to “collect data” seems incredibly stupid to me.

Imagine you are in a casino that doesn’t charge rake. The casino is thinking about charging rake so it turns on the rake on a few tables. What happens?

Everyone will leave the newly raked tables to go to the non raked tables. The casino concludes that they should avoid rake because everyone would stop playing if there was a rake. But this is wrong because at casinos that have every table raked people still play.

TLDR the data that comes from “turning on a few pools to test” will be completely useless and lead to false conclusions bc of spillover effects. We know that exchanges can exist with rake, just add it on everything.

Posting a reply to my tweet since it was also relevant:

Mihai: or add it to all pools of a certain pair, so there’s no rake-less pool to go to (on uniswap)

My reply:

Yes this is good but there is still the issue of bypassing since Uniswap is a network
For example, if you rake DAI ETH, you can go DAI → USDC via DAI/USDC 0.01% fee their and then USDC → ETH via USDC/ETH so you bypass the rake for only a 0.01% fee.

This is particularly a problem for orders that are routed using a smart router, since you will see flow tend to bypass the pool which creates a vicious cycle that can basically make the pool obsolete.

no hard feelings towards @guil-lambert, OP, or anyone else who is advocating for this approach but IMO this is completely the wrong way to go about this. The data we gather will be incredibly misleading.

3 Likes